Affinity columns for the separation of rabbit antibodies (Abs) to the purified nicotinic acetylcholine receptor (AcChR) from Torpedo californica electroplax were constructed in two ways: (1) by direct cross-linking of purified AcChR to cyanogen bromide activated Sepharose 4B, and (2) by first cross-linking a purified curarimimetic neurotoxin to cyanogen bromide activated Sepharose 4B, subsequently saturating with AcChR, and finally cross-linking the toxin to noncovalently bound receptors with bisimidate cross-linkers such as dimethyl suberimidate (DMS). Ab(AcChR) from pooled rabbit antisera were chromatographed in near stoichiometric proportions to the AcChR bound to the column to allow equilibrium selection of antibodies directed against sites which were not sterically hindered. The concept that columns of the second type subfractionate Ab(AcChR) was tested by analyzing Ab(AcChR) from column fractions with an ELISA and radioimmunoassay (RIA) procedure. The ELISA was constructed to that the exposed face of the AcChR was the same as that expected for columns of the second type, i.e. for the internal or cytoplasmic face. Enhanced ratios of ELISA to RIA measures of Ab(AcChR) reflected substantial purification of cytoplasmic face antibodies in DMS-cross linked columns. Total Ab(AcChR) was measured by RIA. Both types of columns gave substantial purification of tightly bound antibodies, i.e. those which were eluted with 3M potassium thiocyanate. Thirty-fifty percent of the IgG eluted was found in these fractions, which contained 2-6% of the total eluted protein. Approximately half of the total IgG present in these fractions represented specific IgG against AcChR. Both types of columns could be reutilized giving similar results; however, their efficiency was diminished.